DataDam Personal Data MCP Server

DataDam is a Model Context Protocol (MCP) server backed by Supabase. It supports both streamable HTTP endpoints and stdio connections, allowing multiple AI tools to share a single personal database.

Important: There is no auth yet. Do not store sensitive data. OAuth is planned.

🚀 Quick Links

• Quickstart Setup - Get started in minutes
• Client Configuration - Connect your AI tool (Claude, ChatGPT, Cursor, etc.)
• Tool Details - Learn how to use each tool
• Troubleshooting - Common issues and solutions

---

How to Use DataDam

The Problem: Your AI forgets everything between conversations. You waste 10-20 minutes every time re-explaining context that should already be known.

The Solution: DataDam is a persistent memory layer that decouples your personal information from your AI tool's unstable memory. Mention something once, and it's remembered forever across all conversations.

How It Works

Just talk naturally - DataDam works invisibly in the background, consulting you for consent:

1. First mention: "I'm using TypeScript with Express for my API project" → DataDam proactively stores this in digital_products

2. Weeks later: "Help me debug this API error" → DataDam retrieves your tech stack (TypeScript, Express) automatically

3. No commands needed - Your AI handles all the storage and retrieval

💡 Tip: If a datapoint you mentioned doesn't save automatically, explicitly tell the AI to save it (e.g., "save that to my interests"). The AI's ability to proactively capture information will improve over time.

---

Example 1: Book Recommendations

❌ WITHOUT DataDam:

You: "Recommend a book for me?"

AI: "What genres do you like? What have you read? Any favorite authors?"

You: "I like sci-fi, I've read Project Hail Mary, The Martian, Children of Time,
     Ender's Game... I prefer hard sci-fi with good character development, nothing
     too violent, I read about 50 books a year so I've probably read most
     popular ones..."

[15 minutes of typing the same preferences you've typed 20 times before]

✅ WITH DataDam:

You: "Recommend a book for me?"

AI: [Automatically retrieves: 50+ books you've read, favorite genres (hard sci-fi),
    preferred themes (character development), dislikes (non-fiction books, violence), reading pace]

    "Based on your love of hard sci-fi like Project Hail Mary and your preference
    for character-driven stories, try 'A Memory Called Empire' by Arkady Martine.
    It's got the political intrigue of Dune with deep character development,
    and you haven't read it yet."

You: [0 minutes of setup. Instant, personalized answer.]

Time saved: 15 minutes → 0 minutes

---

Example 2: Tech Stack Recommendations

❌ WITHOUT DataDam:

You: "How should I add authentication to my API?"

AI: "I can help! What's your tech stack? What database are you using?
     What framework? Are you using TypeScript or JavaScript?"

You: "I'm using Node.js with Express and TypeScript, Supabase for the database
     which is PostgreSQL, I deploy on Render, I use Zod for validation,
     I prefer minimal dependencies, I'm working on an MCP server..."

[15 minutes re-explaining your entire tech stack for the hundredth time]

✅ WITH DataDam:

You: "How should I add authentication to my API?"

AI: [Automatically retrieves: TypeScript, Node.js, Express, Supabase (PostgreSQL),
    Render deployment, Zod validation, MCP development focus]

    "For your Express + TypeScript + Supabase stack, I recommend:
    - Use Supabase Auth (already integrated with your PostgreSQL database)
    - Add a middleware with Zod validation for JWT tokens
    - Keep it minimal (matches your preference)
    - Deploy config already works with your Render setup

    Here's the implementation..."

You: [0 minutes explaining. Instant, stack-specific solution.]

Time saved: 15 minutes → 0 minutes

---

The Core Difference

| Without DataDam | With DataDam | |----------------|--------------| | ❌ Repeat yourself every conversation | ✅ Mention once, remember forever | | ❌ 10-20 min setup per conversation | ✅ 0 min setup, instant context | | ❌ Memory resets when chat resets | ✅ Persistent memory across all chats | | ❌ Context lost when switching AI tools | ✅ Same context across all AI tools | | ❌ Generic, one-size-fits-all responses | ✅ Deeply personalized responses |

---

Available Categories

• contacts - People, emails, relationships
• books - Reading history and preferences
• interests - Hobbies, skills, topics you care about
• digital_products - Tools and technologies you use
• basic_information - Personal details and background
• favorite_authors - Authors you love

This is based off of personal interests and use, feel free to suggest categories to add.

---

Tools Overview

• How it works

  • Your AI tool will invoke the neccessary tools in your console/command line upon needing personal information.
  • It will also fill out the parameters of the call itself
  • Categories group related records (e.g., books, contacts, basic_information). All datapoints are assigned to a category.
  • Tags are used as an optional refinement to narrow down results within each category
  • More information on how each tool works can be found here

• Data model

  • Categories are maintained in the database and surfaced via the data://categories resource, which are static at the moment.
  • Filtering order: choose a category first, then use tags to further narrow results within that category (tags are optional refinements, not replacements).

• Server tools (at …/mcp)

| Tool | Title | Purpose | Required | Optional | | --- | --- | --- | --- | --- | | datadam_search_personal_data | Search Personal Data | Find records by title and content; filter by categories/tags. | query | categories, tags, classification, limit, userId | | datadam_extract_personal_data | Extract Personal Data by Category | List items in one category, optionally filtered by tags. | category | tags, limit, offset, userId, filters | | datadam_create_personal_data | Create Personal Data | Store a new record with category, title, and JSON content. | category, title, content | tags, classification, userId | | datadam_update_personal_data | Update Personal Data | Update fields on an existing record by ID. | recordId | title, content, tags, category, classification | | datadam_delete_personal_data | Delete Personal Data | Delete one or more records; optional hard delete. | recordIds | hardDelete |

• ChatGPT endpoint tools (at …/chatgpt_mcp)

| Tool | Title | Purpose | Required | Optional | | --- | --- | --- | --- | --- | | search | Search (ChatGPT) | Return citation-friendly results for a query. | query | — | | fetch | Fetch (ChatGPT) | Return full document content by ID. | id | — |

Connection Types

DataDam supports two connection methods:

HTTP (Streamable)

• Use case: Hosted deployments, multiple clients, web-based AI tools
• Setup: Deploy to cloud service (e.g., Render), configure clients with URL
• Environment: Server-side environment variables in hosting platform
• Protocol: HTTP/HTTPS with MCP over streamable transport

Stdio (Standard Input/Output)

• Use case: Local development, single-client setups, desktop AI applications
• Setup: Run server.js locally, configure clients to launch the process
• Environment: Local environment variables or passed via client config
• Protocol: MCP over stdio transport with direct process communication

Prerequisites

• Homebrew: Package Manager for MacOS and Linux - Homebrew
• Git: Version control system - Download Git
• Node.js + npm: JavaScript runtime and package manager - Download Node.js
• Accounts: Supabase (required), Render (for hosting)

Quickstart

Installation

1. Clone this repository:

git clone https://github.com/KennethLeeJE8/datadam_mcp.git && cd datadam_mcp

2. Install dependencies:

npm install

3. Build the TypeScript code:

npm run build

Happy to help if you have any problems w the setup! Shoot me a message or send me an email at kennethleeje8@gmail.com :)

Supabase Setup

1. Create a Supabase account

• Go to Supabase Sign Up to create your account
• Important: Remember your password - you'll need it for the database connection later
• Create a new project and wait for it to finish setting up

2. Load the database schema in Supabase SQL Editor:

• Copy the entire contents of src/database/schema.sql
• Supabase Dashboard → SQL Editor → New query
• Paste the copied schema code into the editor
• Click "Run" to execute the schema

3. You should see your Supabase table editor view populated with tables in "Table Editor".

✅ Supabase setup is complete! Your database is ready to use.

Choose Your Connection Type

Select the connection method based on your AI tools and subscription tiers:

• Option A: Stdio (Standard Input/Output)

  • Use for: Coding agents (Cursor, Windsurf, etc.), Claude Desktop (Free tier)
  • Next step: Continue to Local Testing section below

• Option B: HTTP Streamable

  • Use for: ChatGPT Plus or higher, Claude Pro or higher, Coding agents (Cursor, Windsurf, etc.)
  • Next step: Skip to Render Deployment section

Local Testing

1. Set up environment variables by cloning the .env file:

cp .env.example .env

Edit .env and add your Supabase credentials:

To find your SUPABASE_URL:

• Supabase Dashboard → Project Settings → Data API → Project URL

To find your SUPABASE_SERVICE_ROLE_KEY:

• Supabase Dashboard → Project Settings → API Keys → service_role (click "Reveal" to copy)

2. Test the connection with the MCP Inspector:

npm run inspector:stdio

• Transport: Select "stdio"
• Arguments: Enter "server.js"
• Click "Connect"

3. Verify the setup:

• Verify: The inspector should connect and show available tools, confirming Supabase database connection
• Test: Go to the Tools tab and click "List Tools" → find "extract_personal_data_tool" → enter "interests" for categories → click "Run Tool" to verify database connectivity
• You should see a datapoint on "MCP (Model Context Protocol)"

Render Deployment (Only for Streamable HTTP Server)

Feel free to use any hosting platform, this is personal preference.

1. Create a Render account at Render or sign in if you have an existing account

2.

You'll be prompted to fill in the required environment variables:

• Ensure that branch is main
• SUPABASE_URL - Get from: Supabase Dashboard → Project Settings → API → Project URL
• SUPABASE_SERVICE_ROLE_KEY - Get from: Supabase Dashboard → Project Settings → API → Project API keys → service_role (click "Reveal" to copy)

Ensure that the environment variables are filled out correctly, otherwise the deployment will fail.

Verify HTTP Connections

• Health endpoint: curl http://{render_url}/health

Verify HTTP Tools

• Test: Go back to the Command Line and run:

  npm run inspector:http
  

  • Server URL: Enter http://<YOUR_RENDER_URL>/mcp
  • **T